Metraipolis › archive › dc_32fed273e1b9

From Reasoning Strings to Partial Orders: Verifier-Certified Rule Transport through Quotient Policy Optimization — arXiv:2609.27833

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.27833
License (as archived)
CC0-1.0
Archived
2026-10-08T10:57:22.416Z
SHA-256 of the archived text
2a2578b1fb38eb154dd58a839aeb40437cfeed6e56474a9e3935c8c90671421a
Stable URL of this page
https://metraipolis.com/documento/dc_32fed273e1b9

How to cite

From Reasoning Strings to Partial Orders: Verifier-Certified Rule Transport through Quotient Policy Optimization — arXiv:2609.27833 — arXiv (en), https://arxiv.org/abs/2609.27833 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_32fed273e1b9 (Metraipolis archive dc_32fed273e1b9, sha256 2a2578b1fb38eb154dd58a839aeb40437cfeed6e56474a9e3935c8c90671421a)
BibTeX
@misc{dc32fed273e1b9,
  title = {From Reasoning Strings to Partial Orders: Verifier-Certified Rule Transport through Quotient Policy Optimization — arXiv:2609.27833},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.27833},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_32fed273e1b9; sha256 2a2578b1fb38eb154dd58a839aeb40437cfeed6e56474a9e3935c8c90671421a},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.27833. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Many computations admit several valid execution orders because independent subgoals or disjoint state updates can commute. Reinforcement learning with verifiable rewards usually treats each successful trace as a separate token sequence, so serialization choices can be mistaken for logical dependencies. We introduce Verifier-Certified Rule Transport (VCRT), which replays adjacent operation pairs with native verifiers. Pairs whose two orders are accepted and reach the same canonical state provide commutation certificates; rejected or state-changing reversals provide anti-diamonds. VCRT uses anti-diamonds to preserve genuine prerequisites and assigns policy credit to the total probability mass of each certified orbit. It also constrains post-swap consistency, source retention, and policy drift. We evaluate leave-one-environment-out transfer across ProofWriter, CLRS, and Lean through a shared anonymized relation-graph interface. All training and checkpoint decisions are frozen before held-out evaluation, which uses one greedy trajectory per item without search or verifier feedback. VCRT obtains a 77.60% macro pass rate versus 64.53% for the strongest matched baseline, a paired gain of 13.06 points (95% bootstrap CI [12.58, 13.54]). Lean accounts for most of this gain at 33.49 points, while ProofWriter and CLRS improve by 2.85 points on average. Mechanism tests consistently favor anti-diamond supervision, whereas No-Orbit is statistically indistinguishable from full VCRT. The evidence does not establish a general benefit from exact orbit aggregation.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-24 in cs.LG).
Categorie arXiv: Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2609.27833 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-08T10:06:16.750Z dalla pagina https://arxiv.org/catchup/cs.LG/2026-09-24?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).